HP 34401A hacks and upgrades

[Kleinstein]

The measurement of a reference give the combination of noise from the external reference, the meter internal reference (ADR1399 AFAIK) and the ADC noise and gain fluctuations (e.g. from temperature).
I am surprized to see so much difference between the ADR1001 and measurement with the battery, as I would expect the ADR1399 to be more noisy than the ADR1001. The stability of the meters temperature in combination with the meter TC could also make a difference.

With a stable temperature the battery noise should be pretty low for quite some time scale. With the meter one would more see the limit of the internal reference (the 34401 ADC noise with AZ mode is pretty much white noise and thus no issue for longer times). To really test battery noise one would need to check the difference between 2 battery packs.

With classic Li-Ion cells the TC depends on the charge state / voltage, and it incudes TVC zero crossing at some 3.6 V for the cells that I tested.
I don't know the Sodium chells, but the charge state may also make a difference.
Another point is that the cells need quite some time (e.g. several days) to stabilitze after charge / discharge.

[iMo]

The sim of above transistorised bootstrapped opamp in some 34401A boxes based on the shot from bbs.38hot.net above (replay #273).

Picky.. Sometimes oscillates in the sim.. Would be great to see a better readable schematics.
It did not work with the R5 value 12k in the shot. It should be 1k2 imho.
Also the R131 in the original shot is not 12k but 1k2 most probably (it sets the current source such it works nicely, you have to play with it).

Input 24Vpp sine, 100Hz OPA145, the zener is 9.1V in my case.

Mod for upcoming 34402:

Added a square 36Vpp inp OPA145

Added a square 36Vpp inp with ADA4528 (5V RRIO AZ), 5V1 zener

And how to set the Vcc_op/Vee_op symmetry and the current source..

[iMo]

I am surprized to see so much difference between the ADR1001 and measurement with the battery, as I would expect the ADR1399 to be more noisy than the ADR1001..

That is something to investigate. Some people say the 10V output version of the ADR1001 is pretty noisy because of the noisy dividers there (the internal 6.6V->5V and 5V->10V dividers are quite noisy). Also people say the reference itself is good (LTZ level).
One maker here in blog is using only the 6.6V output, bypassing the internal dividers and the buffer, if I understood him correctly.

[mawyatt]

The above transistorised bootstrapped opamp in the 34401A based on the shot from bbs.38hot.net above.
Picky.. Sometimes oscillates.. Would be great to see a better readable schematics.
It did not work with the R5 value 12k in the shot. It should be 1k2 imho.
Zener is 9.1V in my case.
Input 24Vpp sine, 100Hz.

PS: added a +/- 20V input version..
Input 40Vpp sine, 100Hz.

added a square 36Vpp inp sim

Don't think Q5 is connected correctly, it's operating inverted!!

Best

[GigaJoe]

R131 == R5 ? Then 1211
https://imgtree.co/direct/313PAfMe.webp
https://imgtree.co/direct/4wIDOcf2.webp
https://imgtree.co/direct/n4No5nFt.webp

?:   are u trying to replicate it to fit something in a bit over 36V rails ?

[iMo]

This is with values closest to the original one from the pcb.
The voltage at the opamp is not perfectly symmetrical, needs be finetuned as depicted in my previous sim.
Also the sim does not work with such small current as set in original, like 1.7mA.
So perhaps the opamp model is not fit for such sim, or the zener.
And it could be there is a wiring error somewhere.
Otherwise it works the same as in my previous sims, it seems.

?:   are u trying to replicate it to fit something in a bit over 36V rails ?

Yep, for +/-20V input (needs +/-24V Vcc/Vee) - see above sims.. Not for 34401P, of course..
See below the sim with 40Vpp square and ADA4528 opamp (an example only). OPA140 or OPA145 would be nicer there as it has small inp current (and it is not AZ). Plus some BAV199s around its input.
The difference Vout-Vinp for +/-20V is +1.4uV..-2.0uV..

[GigaJoe]

So when I measure my old one, the rails were around 18.7–.9 ish... and I kind of scratch my head about what the intention was to bump the rails a bit over max, and then do the grueling work of overcoming difficulties.

so why it need so high ?  and can it be lowered a bit to accommodate a good modern opamp , without that funky fuss of tracking supply rails ?

[Kleinstein]

The +-18 V supply in the HP34401 is a bit odd, as most parts reduce the supply to +-15 V with local zeners anyway. To be safely on the off side one may want -18 V for JFET switching, in case an amplifier shows ringing to -15 V. Some amplifiers (e.g. OPA140 and related) may want extra headroom to work all the way to +-12 V.

For the bootstrapped OP-amp the supply does not need to be symmetric (or the common mode voltage in the center). It depends on the OP-amp that is used. E.g. the OPA145 has a CM range from 0 to VCC-3.5 V and would thus wants a voltage more to the negative side, especially with a relatively low supply. The original OPA130 is more symmetic an a point close to the center makes sense.

A 34401 class DMM could now get away with a good OP-amp (e.g. OPA140 and related). The somewhat limiting factor is (was) the common mode rejection. Ideally one would keep the load away from the amplifier to avoid thermal effects for best linearity.

For the 34401 the amplifier / buffer is still usually not the limiting factor for the noise.

[iMo]

..With a stable temperature the battery noise should be pretty low for quite some time scale...

Below an experiment wih 100ohm/6.25F RC low pass after the ADR1001#1 10V.

After 2 days soaking the "capacitor" (4x 25F/2.7V in series) had 30nA leakage at 30C ambient and 9.9V.
My rough estimate that leakage with 100ohm should create <200nV/C TC..
Let us see how it will perform..

[Kleinstein]

Similar to batteries the capacitors will also have a temperature coefficient for the voltage. So the capacitors also need a stable temperature.
100 ohms and 6 F would be a 10 minutes time constant. So at least the slower temperature drift would not have much effect.

[iMo]

That leakage above I saw (0.3V @ 10Meg in ser) might be perhaps a nonsense and it indicated some special state of the capacitors during soaking. I do not use balancing there (the capacitors were salvaged from a blade server they were balanced there) so the 30nA looks like full BS. I can see it on the time it needs to make the last mV or so, it takes hours still slowly rising so the leakage will be by orders of magnitude higher. Also the drop there is like 3mV so it consumes around 30uA. Anyhow, I let it charge today from the 10V ref.

[Kleinstein]

Similar to normal electrolytic capacitors I would expect quite some dielectric absorbtion. So it takes a long time for the leakage current to settle.

[iMo]

Btw., could a voltage reference with the output buffer in the form of an upper side npn emitter follower transistor (with a current limiter) sourcing a 10k (or similar value) resistor against the ground work properly with a 6 Farad load wired via 100ohm resistor?
Imho the positive charges are pushed into the "100/6F" low pass, while the negative drops are discharged via "10100/6F" RC (so 100x slower).
Would not it work as a "peak detector" where the voltage on that 6F cap will still creep upwards on the upper envelope of the ref voltage?

[iMo]

Below the result of the ad hoc experiment with 100ohm/6.25F low pass filter after the ADR1001#1 10V Vref.
A full disaster..

UPdated2..

[iMo]

For the record:

In the HP34970A (the same ADC as in the 34401A) they create the +5V_REF for the AUX_ADC (in the main MCU) from a 5V zener (via 2k off +18V rail), decoupled by 10nF..

[cellularmitosis]

Here is a new hack: VFD replacement!

Great to see this thread alive and well after so many years.  Super thanks to iMo for all of your work!

I thought I would drop a link to a little external serial display "hack" I made for the 34401a a few years back:

https://gist.github.com/cellularmitosis/1582236f226e9d98075a0c971eb4168c

[GigaJoe]

AC response 1V range ,  0.5V input,  modded 34401A
i use my basic analog gwinstek gen, so output may float a bit,
here # : ( 10Hz 0.507 , 100Hz 0.509 , 1Khz 0.509 , 10Khz 0.511, 100Khz 0.519, 200Khz 0.524, 300khz 0.528, 400khz 0.510, 500khz 0.502,800khz 0.520, 1Mhz 0.540, 1.4Mhz 0.680, 1.6Mhz 0.854,2.3Mhz 0.505, 3.2Mhz 0.233)
3.2M - limit for this gen.
1.6- a peak,  some kind resonance or something,